Initial: 4 retrieval eval runs on a 996-note Obsidian vault

- 20260618-233912.csv: 7 models, 54 notes, 29 queries (toy slice, saturated)
- 20260619-025330.csv: 7 models, 995 notes, 150 queries (real pile, ranking inverted)
- 20260620-142445-wholenote.csv: 7 models, 996 notes, 450 queries (ranking held)
- 20260620-172625-chunked.csv: 1 model, 18279 chunks, 450 queries (chunking A/B)

Companion to kylebrodeur/embed-eval-on-your-vault and the KE-1 blog post.

20260618-233912.csv

@@ -0,0 +1,8 @@
+name,provider,model,dim,recall@1,recall@5,recall@10,mrr@10,q_latency_ms,approx_cost_usd
+embeddinggemma,ollama,embeddinggemma,768,0.6552,0.9655,1.0,0.8027,18.2,0.0
+bge-m3,ollama,bge-m3,1024,0.6897,1.0,1.0,0.8,31.5,0.0
+bge-large,ollama,bge-large,1024,0.8276,0.931,0.9655,0.8693,29.4,0.0
+nomic-embed-text,ollama,nomic-embed-text,768,0.8276,0.9655,1.0,0.9004,16.1,0.0
+mxbai-embed-large,ollama,mxbai-embed-large,1024,0.7931,0.931,0.9655,0.8487,27.0,0.0
+paraphrase-multilingual,ollama,paraphrase-multilingual,768,0.6552,0.931,1.0,0.7869,16.4,0.0
+all-minilm (current local fallback),ollama,all-minilm,384,0.6552,0.9655,1.0,0.7787,8.3,0.0

20260619-025330.csv

@@ -0,0 +1,8 @@
+name,provider,model,dim,recall@1,recall@5,recall@10,mrr@10,q_latency_ms,approx_cost_usd
+embeddinggemma,ollama,embeddinggemma,768,0.8,0.92,0.9267,0.8524,18.7,0.0
+bge-m3,ollama,bge-m3,1024,0.7533,0.9067,0.9333,0.8172,34.6,0.0
+bge-large,ollama,bge-large,1024,0.7333,0.88,0.9133,0.7984,28.3,0.0
+nomic-embed-text,ollama,nomic-embed-text,768,0.7267,0.86,0.9,0.7912,14.8,0.0
+mxbai-embed-large,ollama,mxbai-embed-large,1024,0.7067,0.8933,0.9067,0.7865,29.5,0.0
+paraphrase-multilingual,ollama,paraphrase-multilingual,768,0.62,0.78,0.8333,0.6903,16.3,0.0
+all-minilm (current local fallback),ollama,all-minilm,384,0.64,0.7933,0.8333,0.6998,6.2,0.0

20260620-142445-wholenote.csv

@@ -0,0 +1,8 @@
+name,provider,model,dim,recall@1,recall@5,recall@10,mrr@10,q_latency_ms,approx_cost_usd
+embeddinggemma,ollama,embeddinggemma,768,0.7089,0.9022,0.94,0.7918,20.2,0.0
+bge-m3,ollama,bge-m3,1024,0.68,0.88,0.92,0.7669,32.0,0.0
+bge-large,ollama,bge-large,1024,0.6111,0.8267,0.8756,0.7056,27.7,0.0
+nomic-embed-text,ollama,nomic-embed-text,768,0.6156,0.8511,0.9,0.716,13.7,0.0
+mxbai-embed-large,ollama,mxbai-embed-large,1024,0.6356,0.8444,0.8956,0.7303,28.0,0.0
+paraphrase-multilingual,ollama,paraphrase-multilingual,768,0.4844,0.7,0.76,0.5773,16.4,0.0
+all-minilm (current local fallback),ollama,all-minilm,384,0.5911,0.7978,0.8489,0.6841,6.5,0.0

20260620-172625-chunked.csv

@@ -0,0 +1,2 @@
+name,provider,model,dim,recall@1,recall@5,recall@10,mrr@10,q_latency_ms,approx_cost_usd
+embeddinggemma,modal,embeddinggemma,768,0.6533,0.8889,0.9356,0.7524,6.7,0.0

README.md

@@ -0,0 +1,73 @@
+---
+license: mit
+task_categories:
+  - feature-extraction
+  - retrieval
+tags:
+  - embeddings
+  - retrieval
+  - rag
+  - benchmarks
+  - embedding-models
+size_categories:
+  - n<1K
+pretty_name: "Embedding Eval Results — Round 1, Stage 1, Chunked A/B"
+---
+
+# Embedding Eval Results
+
+The committed results from a real embedding-model benchmark that embarrassed a leaderboard's recommendation.
+
+## What's in here
+
+Four CSV files representing four evaluation runs on a personal Obsidian vault:
+
+| File | Notes | Queries | Models | Purpose |
+|---|---|---|---|---|
+| `20260618-233912.csv` | 54 | 29 | 7 | Round 1 — toy slice. **Saturated benchmark.** |
+| `20260619-025330.csv` | 995 | 150 | 7 | Stage 1 — real pile. The ranking inverted. |
+| `20260620-142445-wholenote.csv` | 996 | 450 | 7 | Stage 1 enlarged — held the new order. |
+| `20260620-172625-chunked.csv` | 18,279 chunks | 450 | 1 | Chunked A/B — chunking hurt every metric. |
+
+## What the data shows
+
+Round 1 on 54 notes: all 7 models looked great. recall@5 above 90% for every candidate. The "winner" by leaderboard-style ranking was `nomic-embed-text`.
+
+Stage 1 on 995 notes: the ranking inverted. The mid-pack model `embeddinggemma` took first on every metric. The round-one "winner" sank to mid-pack. The default the system had been quietly using, `all-minilm`, dropped to last.
+
+Same code. Same models. Same scoring. The only thing that changed was the size of the pile.
+
+## Columns
+
+Each CSV has the same columns:
+
+```
+name,provider,model,dim,recall@1,recall@5,recall@10,mrr@10,q_latency_ms,approx_cost_usd
+```
+
+- `name`: human-readable label (matches the model except for the legacy `all-minilm` row, which is annotated)
+- `provider`: `ollama` (local) or `modal` (the project's own GPU service) — see sources below
+- `model`: the actual model name passed to the embedder
+- `dim`: embedding dimension
+- `recall@1, recall@5, recall@10`: fraction of queries where the source note landed in the top-k
+- `mrr@10`: mean reciprocal rank
+- `q_latency_ms`: query-side embedding latency (note embedding happens separately)
+- `approx_cost_usd`: zero for local models; the cheapest paid model cost roughly a few cents per run
+
+## Source
+
+The full eval harness is in the [pi-vault-mind](https://github.com/kylebrodeur/pi-vault-mind) repo at `eval/run_eval.py`. The companion blog post is ["Pick the Model From Your Own Data"](https://kylebrodeur.substack.com) (KE-1).
+
+The simplified standalone version of the harness, written for the blog post, is at [kylebrodeur/embed-eval-on-your-vault](https://github.com/kylebrodeur/embed-eval-on-your-vault). It is a single-file Python script with no dependencies beyond `python3` that runs against any vault of `.md` or `.txt` notes.
+
+## Reproducibility
+
+Each CSV is timestamped (UTC). The harness commits the result on every run. Re-running with the same model/provider/dataset should produce the same numbers within rounding. Embedding models do drift across versions, so if you re-run after a model bump, expect ±1-2% on the metrics.
+
+## Citation
+
+If you use these numbers in another piece, cite the model row and the run timestamp. The full provenance is in the [pi-vault-mind](https://github.com/kylebrodeur/pi-vault-mind) repo commit log (the 2026-06-18 to 2026-06-20 range).
+
+## License
+
+MIT. Numbers are facts; the framing is the author's.